Re: [PATCH v9 00/29] drm/msm/dp: Add MST support for MSM chipsets

From: Joonhoe Kim

Date: Thu Nov 19 2026 - 06:54:26 EST


On Thu, 19 Nov 2026, Yongxing Mou wrote:
> Add support for Multi-stream transport for MSM chipsets that allow
> a single instance of DP controller to send multiple streams.

Hi Yongxing,

On v7 I reported that s2idle hangs on resume with MST active, and you
said I could send the fix. I will not have time to follow it up as a
separate patch, so here it is for you to take into the series if it is
useful.

The series never calls drm_dp_mst_topology_mgr_suspend() or
drm_dp_mst_topology_mgr_resume(). While MST is active the link is kept
up after the last stream is disabled, so system suspend powers the
controller down with link_ready still set; on resume the first stream
enable skips link training, and the payload table write hangs retrying
AUX over a link that is down.

The fix suspends the topology manager and takes the link down before the
controller is suspended, and on resume initializes the PHY again and
resumes the topology manager (turning MST off with a hotplug event if
the branch device is gone). The version on top of v7 of this series is
here:

https://github.com/joonhoekim/tb323fu-linux/blob/0e73bc96d65386bee31cc003f2e76a2e9623f631/kernel/patches/0202-drm-msm-dp-suspend-and-resume-the-MST-topology-acros.patch

With it, a Lenovo Y700 Gen 5 (TB323FU, Kaanapali) with two displays on
an MST hub resumed from s2idle 3 times out of 3. On v9,
msm_dp_display_unprepare() should move into the patch's
msm_dp_mst_suspend() and run under mst_lock, as
msm_dp_mst_stream_post_disable() holds that lock when it calls it. If
post_disable starts taking the link down after the last stream
(Sashiko's comment on 18/29), only the topology manager suspend/resume
part is still needed. Feel free to take it under your own name or keep
my authorship.

Thanks,
Joonhoe Kim